Many thanks!
I will try this.

Best,
Maria

________________________________________
Lähettäjä: FSL - FMRIB's Software Library <[log in to unmask]> käyttäjän  puolestaMark Jenkinson <[log in to unmask]>
Lähetetty: 6. tammikuuta 2017 10:49
Vastaanottaja: [log in to unmask]
Aihe: Re: [FSL] Wrong labels in FSL

Dear Maria,

It is the qform and sform matrix that determine everything with respect to the labels.
The information in the one you included in your email is not OK though, as both qform_code and sform_code are zero and that indicates that the contents can be ignored, and hence the labels will revert to the defaults in FSLView and not use the information in the header.

If you are setting things up for yourself then the qform is awkward to deal with so I would recommend setting the sform_code to 1 and then setting the values in the sform matrix so that they correctly specify the directions of the main axes (the matrix maps from the voxel coordinate system to the real world coordinate system where, in the real world coordinate system, x=left-right, y=posterior-anterior and z=inferior-superior).  Also make sure that the voxel size is captured by the sform (i.e. if you move one voxel unit in a direction that the step size in the real world, mm, coordinate system is of the appropriate size - this is easily done by using the mm voxel sizes in the appropriate rows).

You may be able to easily work out what you need from first principles, or you might find it easier to take a trial and error approach.  Either is fine as long as your labels are correct in FSLView at the end of the process. *However* there is a great danger that left and right can become confused in this.  If you do not have the information from the original scanner as to what orientation the voxels are in wrt the real left-right axis then you might easily decide on something that "looks" correct but is flipped in the left-right axis. This is a real problem with doing your own reconstructions, versus using standard tools, as  there is usually no way to safely look at a scan and determine this from the anatomy (unless you have a known, asymmetric pathology or anatomical variant).  If you had a structural scan that had been reconstructed by a standard tool and you were confident in its left-right labelling, then you should be able to use that to visually identify in the EPI which side is left or right by comparing the angle of the head, or variations in the inter-hemispherical fissure (especially at the posterior end), or by careful inspection of the folding patterns (e.g. in the cingulate). Once you've identified visually what is the correct left and right then you can check whether your modified sform is giving you the right labelling.  If you do not have any scans that are converted by standard tools, or cannot view the anatomy of the scanner console or with a dicom viewer or similar, then this does become a real problem for determining left and right, and you'll have to find some way of deciding this for yourself.

I hope this helps.
All the best,
        Mark



> On 5 Jan 2017, at 10:29, Maria Hakonen <[log in to unmask]> wrote:
>
> Dear FSL experts,
>
> Could you please let me know which information is critical to get the labels (i.e. right, left, anterior, posterior...) correctly in fslview?  I have reconstructed raw fMRI data using a Matlab script that gives the data in .mat format. I have converted the data to the nifti format using NIFTI toolbox but some of the header information is missing / incorrect.
>
> An fMRI data that is measured with standard EPI and that also has the labels correctly in fslview has the following fields in the header:
> hdr.hist
>
> ans =
>
>        descrip: 'EPI 64 5_5_4_3min'
>       aux_file: ''
>     qform_code: 0
>     sform_code: 0
>      quatern_b: 0
>      quatern_c: 1
>      quatern_d: 0
>      qoffset_x: 125.4668
>      qoffset_y: -165.0882
>      qoffset_z: -10.3035
>         srow_x: [1 0 0 0]
>         srow_y: [0 1 0 0]
>         srow_z: [0 0 1 0]
>    intent_name: ''
>          magic: 'n+1'
>     originator: [22.9066 34.0176 3.5759 0 -32768]
>     rot_orient: [1 2 3]
>    flip_orient: [3 0 0]
>
> Should the labels be shown correctly in fslview if I manage to find the corresponding information and create the similar struct for my data as standard EPI measurements have.
>
> I would really appreciate any help you can provide!
>
> Best regards,
> Maria